Reconstructing Neutrino Energy using CNNs for GeV Scale IceCube Events

The IceCube Collaboration

Reconstructing Neutrino Energy using CNNs for GeV Scale IceCube Events

The IceCube Collaboration

Physics and Astronomy

Michigan State University
Loyola University Chicago
German Electron Synchrotron
University of Canterbury
Université libre de Bruxelles
University of Copenhagen
Oskar Klein Centre
University of Geneva
Karlsruhe Institute of Technology
University of Delaware
Harvard University
Marquette University
Pennsylvania State University
Friedrich-Alexander University Erlangen-Nürnberg
University of Wisconsin-Madison
Massachusetts Institute of Technology
South Dakota School of Mines & Technology
University of California at Irvine
University of California at Berkeley
Ohio State University
University of Wuppertal
Ruhr University Bochum
Technical University of Munich
University of Rochester
University of Maryland, College Park
University of Padua
University of Kansas
Moscow Engineering Physics Institute
Lawrence Berkeley National Laboratory
RWTH Aachen University
Johannes Gutenberg University Mainz
Uppsala University

Research output: Contribution to journal › Conference article › peer-review

1 Scopus citations

Abstract

Measurements of neutrinos at and below 10 GeV provide unique constraints of neutrino oscillation parameters as well as probes of potential Non-Standard Interactions (NSI). The IceCube Neutrino Observatory’s DeepCore array is designed to detect neutrinos down to GeV energies. IceCube has built the world’s largest data set of neutrinos >10 GeV, making searches for NSI a computational challenge. This work describes the use of convolutional neural networks (CNNs) to improve the energy reconstruction resolution and speed of reconstructing O(10 GeV) neutrino events in IceCube. Compared to current likelihood-based methods which take seconds to minutes, the CNN is expected to provide approximately a factor of 2 improvement in energy resolution while reducing the reconstruction time per event to milliseconds, which is essential for processing large datasets.

Original language	English
Article number	1053
Journal	Proceedings of Science
Volume	395
State	Published - Mar 18 2022
Event	37th International Cosmic Ray Conference, ICRC 2021 - Virtual, Berlin, Germany Duration: Jul 12 2021 → Jul 23 2021

Cite this

@article{ccb017735dbe4bf08311ccdcce20b096,

title = "Reconstructing Neutrino Energy using CNNs for GeV Scale IceCube Events",

abstract = "Measurements of neutrinos at and below 10 GeV provide unique constraints of neutrino oscillation parameters as well as probes of potential Non-Standard Interactions (NSI). The IceCube Neutrino Observatory{\textquoteright}s DeepCore array is designed to detect neutrinos down to GeV energies. IceCube has built the world{\textquoteright}s largest data set of neutrinos >10 GeV, making searches for NSI a computational challenge. This work describes the use of convolutional neural networks (CNNs) to improve the energy reconstruction resolution and speed of reconstructing O(10 GeV) neutrino events in IceCube. Compared to current likelihood-based methods which take seconds to minutes, the CNN is expected to provide approximately a factor of 2 improvement in energy resolution while reducing the reconstruction time per event to milliseconds, which is essential for processing large datasets.",

author = "\{The IceCube Collaboration\} and Jessie Micallef and R. Abbasi and M. Ackermann and J. Adams and Aguilar, \{J. A.\} and M. Ahlers and M. Ahrens and C. Alispach and Alves, \{A. A.\} and Amin, \{N. M.\} and R. An and K. Andeen and T. Anderson and G. Anton and C. Arg{\"u}elles and Y. Ashida and S. Axani and X. Bai and Balagopal, \{A. V.\} and A. Barbano and Barwick, \{S. W.\} and B. Bastian and V. Basu and S. Baur and R. Bay and Beatty, \{J. J.\} and Becker, \{K. H.\} and \{Becker Tjus\}, J. and C. Bellenghi and S. BenZvi and D. Berley and E. Bernardini and Besson, \{D. Z.\} and G. Binder and D. Bindig and E. Blaufuss and S. Blot and M. Boddenberg and F. Bontempo and J. Borowka and S. B{\"o}ser and O. Botner and J. B{\"o}ttcher and E. Bourbeau and F. Bradascio and J. Braun and S. Bron and J. Brostean-Kaiser and S. Browne and J. Kiryluk",

note = "Publisher Copyright: {\textcopyright} Copyright owned by the author(s).; 37th International Cosmic Ray Conference, ICRC 2021 ; Conference date: 12-07-2021 Through 23-07-2021",

year = "2022",

month = mar,

day = "18",

language = "English",

volume = "395",

journal = "Proceedings of Science",

issn = "1824-8039",

}

TY - JOUR

T1 - Reconstructing Neutrino Energy using CNNs for GeV Scale IceCube Events

AU - The IceCube Collaboration

AU - Micallef, Jessie

AU - Abbasi, R.

AU - Ackermann, M.

AU - Adams, J.

AU - Aguilar, J. A.

AU - Ahlers, M.

AU - Ahrens, M.

AU - Alispach, C.

AU - Alves, A. A.

AU - Amin, N. M.

AU - An, R.

AU - Andeen, K.

AU - Anderson, T.

AU - Anton, G.

AU - Argüelles, C.

AU - Ashida, Y.

AU - Axani, S.

AU - Bai, X.

AU - Balagopal, A. V.

AU - Barbano, A.

AU - Barwick, S. W.

AU - Bastian, B.

AU - Basu, V.

AU - Baur, S.

AU - Bay, R.

AU - Beatty, J. J.

AU - Becker, K. H.

AU - Becker Tjus, J.

AU - Bellenghi, C.

AU - BenZvi, S.

AU - Berley, D.

AU - Bernardini, E.

AU - Besson, D. Z.

AU - Binder, G.

AU - Bindig, D.

AU - Blaufuss, E.

AU - Blot, S.

AU - Boddenberg, M.

AU - Bontempo, F.

AU - Borowka, J.

AU - Böser, S.

AU - Botner, O.

AU - Böttcher, J.

AU - Bourbeau, E.

AU - Bradascio, F.

AU - Braun, J.

AU - Bron, S.

AU - Brostean-Kaiser, J.

AU - Browne, S.

AU - Kiryluk, J.

PY - 2022/3/18

Y1 - 2022/3/18

N2 - Measurements of neutrinos at and below 10 GeV provide unique constraints of neutrino oscillation parameters as well as probes of potential Non-Standard Interactions (NSI). The IceCube Neutrino Observatory’s DeepCore array is designed to detect neutrinos down to GeV energies. IceCube has built the world’s largest data set of neutrinos >10 GeV, making searches for NSI a computational challenge. This work describes the use of convolutional neural networks (CNNs) to improve the energy reconstruction resolution and speed of reconstructing O(10 GeV) neutrino events in IceCube. Compared to current likelihood-based methods which take seconds to minutes, the CNN is expected to provide approximately a factor of 2 improvement in energy resolution while reducing the reconstruction time per event to milliseconds, which is essential for processing large datasets.

AB - Measurements of neutrinos at and below 10 GeV provide unique constraints of neutrino oscillation parameters as well as probes of potential Non-Standard Interactions (NSI). The IceCube Neutrino Observatory’s DeepCore array is designed to detect neutrinos down to GeV energies. IceCube has built the world’s largest data set of neutrinos >10 GeV, making searches for NSI a computational challenge. This work describes the use of convolutional neural networks (CNNs) to improve the energy reconstruction resolution and speed of reconstructing O(10 GeV) neutrino events in IceCube. Compared to current likelihood-based methods which take seconds to minutes, the CNN is expected to provide approximately a factor of 2 improvement in energy resolution while reducing the reconstruction time per event to milliseconds, which is essential for processing large datasets.

UR - https://www.scopus.com/pages/publications/85122867451

M3 - Conference article

AN - SCOPUS:85122867451

SN - 1824-8039

VL - 395

JO - Proceedings of Science

JF - Proceedings of Science

M1 - 1053

T2 - 37th International Cosmic Ray Conference, ICRC 2021

Y2 - 12 July 2021 through 23 July 2021

ER -

Reconstructing Neutrino Energy using CNNs for GeV Scale IceCube Events

Abstract

Other files and links

Fingerprint

Cite this